Applying B-tagging efficiencies independently?

[Jim Hirschauer]

In our Delphes cards:
​http://cmssw.cvs.cern.ch/cgi-bin/cmssw.cgi/UserCode/spadhi/Snowmass/Cards/delphes_card_Snowmass_140PileUp.tcl?revision=1.5&view=markup
we have two b-tagging working points (BTagging and BTaggingLoose).

If I understand how to correctly access the b-tagging bits, I observe that a given jet can pass the BTagging tagger but not the BTaggingLoose tagger, and vice versa.

This would be expected behavior if the two working points were independent, but in cases that the working points correspond to different cuts on a single continuous discriminant, they should not be independent.

I'm not sure this is critical, but I thought I'd mention it.

[Michele Selvaggi]

Hi Jim,

thanks for pointing this out.
The reason for what you see is that the b-tagging module is run twice with different random numbers.

As you point out, running with the same random number would prevent this, but would not cover the case where the working point is derived from different discriminants.

This could eventually become a problem as soon as you are statistically limited and start looking at single events. I don't see this happening with Delphes however.

Anyway, please let us know if this becomes an issue for you.

Michele